Adversarial Bayesian Augmentation for Single-Source Domain Generalization
This addresses the challenge of domain generalization for image classification when training data is limited and target domains are inaccessible, though it appears incremental as it builds on existing augmentation techniques.
The paper tackles the problem of generalizing to unseen image domains in single-source domain generalization by proposing Adversarial Bayesian Augmentation (ABA), which uses adversarial learning and Bayesian neural networks to generate diverse data augmentations, and it outperforms all previous state-of-the-art methods across various domain shifts.
Generalizing to unseen image domains is a challenging problem primarily due to the lack of diverse training data, inaccessible target data, and the large domain shift that may exist in many real-world settings. As such data augmentation is a critical component of domain generalization methods that seek to address this problem. We present Adversarial Bayesian Augmentation (ABA), a novel algorithm that learns to generate image augmentations in the challenging single-source domain generalization setting. ABA draws on the strengths of adversarial learning and Bayesian neural networks to guide the generation of diverse data augmentations -- these synthesized image domains aid the classifier in generalizing to unseen domains. We demonstrate the strength of ABA on several types of domain shift including style shift, subpopulation shift, and shift in the medical imaging setting. ABA outperforms all previous state-of-the-art methods, including pre-specified augmentations, pixel-based and convolutional-based augmentations.